Bacterial promoter modelling and prediction for E. coli and B. subtilis with Beagle

Abstract

We constructed sigma70-promoter models of varying complexity to predict promoter locations and to evaluate the importance of specific promoter elements. For this purpose, a novel software, named Beagle, was developed that utilizes an easy description language to conveniently specify promoter models. Model specifications are translated into position weight matrices and gap distributions which are refined using data
from known promoters. The method is transparent, fast and allows the rapid exploration of different promoter models. Applied to promoter prediction in E. coli and B. subtilis,
we show that inclusion of UP-elements and extended -10 motifs into the model yields a significant increase in prediction accuracy. The software, data sets and extended results can be downloaded at http://eresearch.fit.qut.edu.au/Beagle/.